TONresistor · TenSapphire · Sep 1, 2026
diff --git a/GETTING_STARTED.md b/GETTING_STARTED.md
@@ -98,7 +98,7 @@ You should see:
 ✅ Knowledge indexed
 ✅ Telegram: @your_agent connected
 ✅ TON Blockchain: connected
-✅ Agent is ready! (131 base tools)
+✅ Agent is ready! (102 base tools)
 ```
 
 **Verify:** Send `/ping` to your agent on Telegram.
@@ -169,15 +169,11 @@ Admin commands are only available to users listed in `admin_ids`. All commands w
 
 ## Tool Categories
 
-Teleton has **131 always-registered tools**, plus 5 optional system tools:
+Teleton has **102 always-registered tools**, plus 5 optional system tools:
 
 | Category | Count | Highlights |
 |----------|-------|------------|
-| **Telegram** | 85 | Messaging, media, chats, groups, polls, stickers, gifts, stars, stories, contacts, folders, profile, memory, tasks |
-| **TON & Jettons** | 15 | W5R1 wallet, send/receive TON & jettons, balances, prices, holders, history, charts, NFTs, DEX quotes |
-| **STON.fi DEX** | 5 | Swap, quote, search, trending tokens, liquidity pools |
-| **DeDust DEX** | 5 | Swap, quote, pools, prices, token info |
-| **TON DNS** | 8 | Domain check, auctions, bidding, resolution, TON Sites |
+| **Telegram** | 83 | Messaging, media, chats, groups, polls, stickers, gifts, stars, stories, contacts, folders, profile, memory, tasks |
 | **Journal** | 3 | Log trades/operations with reasoning and P&L |
 | **Workspace** | 6 | Sandboxed file operations |
 | **Web** | 2 | Search and page extraction |
@@ -323,7 +319,7 @@ For contributors, create a TypeScript tool in `src/agent/tools/` and register it
 src/
 ├── index.ts        # Main application entry point (TeletonApp)
 ├── agent/          # LLM runtime and tool registry
-│   └── tools/      # telegram/, ton/, stonfi/, dedust/, dns/, journal/, workspace/
+│   └── tools/      # telegram/, journal/, workspace/, web/, spotify/, exec/
 ├── telegram/       # GramJS bridge, message handlers, admin commands, debouncing
 ├── memory/         # SQLite database, RAG search (FTS5 + vector), compaction
 ├── ton/            # Wallet operations, payment verification, TON blockchain

diff --git a/HUMANIZATION_IMPLEMENTED.md b/HUMANIZATION_IMPLEMENTED.md
@@ -0,0 +1,103 @@
+# 🎭 Реализованные улучшения для более человечного поведения
+
+## ✅ Фаза 1: Variable Typing Delays (ГОТОВО)
+
+### Что сделано
+
+1. **Новый модуль `src/telegram/human-behavior.ts`**
+   - `calculateTypingDelay()` - расчет времени "печатания" на основе длины ответа
+   - `calculateReadDelay()` - задержка перед началом печати (чтение сообщения)
+   - `shouldShowTyping()` - интеллектуальное решение показывать ли typing
+   - `isSimpleAcknowledgment()` - детекция простых ответов типа "ок", "да"
+   - `calculateReactionDelay()` - естественная задержка перед реакцией
+
+2. **Интеграция в message handler** (`src/telegram/handlers.ts`)
+   - Задержка перед началом typing (имитация чтения)
+   - Variable typing delay зависит от длины ответа
+   - Короткие ответы типа "ok" могут пропускать typing (50% шанс)
+   - Группы получают ответы на 40% быстрее (меньше думаем в groups)
+
+3. **Интеграция в heartbeat** (`src/heartbeat.ts`)
+   - Проактивные сообщения тоже используют human-like delays
+   - Fallback на конфиг `reply_delay_ms` если typing отключен
+
+### Параметры по умолчанию
+
+```typescript
+baseDelayMs: 800        // Базовое время "думания"
+charsPerSecond: 25      // Скорость печати (реалистично)
+minDelayMs: 500         // Минимум 0.5с
+maxDelayMs: 8000        // Максимум 8с
+variancePercent: 30     // ±30% рандомности
+
+// Groups: все задержки × 0.6
+```
+
+### Примеры поведения
+
+| Длина ответа | DM задержка | Group задержка | Примечание |
+|--------------|-------------|----------------|------------|
+| 20 chars (короткий) | ~1.2s | ~0.7s | Может пропустить typing |
+| 100 chars | ~2.8s | ~1.7s | Типичный ответ |
+| 500 chars (длинный) | ~7.5s | ~4.5s | Серьёзный ответ |
+| "ok" (acknowledgment) | 0-1.2s | 0-0.7s | 50% без typing |
+
+### Отключение TON функций (БОНУС)
+
+Добавлен конфиг для отключения крипто-инструментов:
+
+```yaml
+# config.yaml
+ton_features:
+  enabled: false  # По умолчанию выключено
+```
+
+Экономия:
+- ❌ Не создается createTonSDK (733 строки, complexity 102)
+- ❌ Не грузятся blockchain зависимости
+- ✅ Плагины получают `sdk.ton = null`
+
+---
+
+## 🚀 Следующие шаги (не реализовано)
+
+### Фаза 2: Context-Aware Behavior
+1. **Smart silence** - молчать в активных групповых дискуссиях
+2. **Variable reply probability** - не отвечать на каждое сообщение в группах
+3. **Time-of-day awareness** - стиль общения зависит от времени суток
+4. **Conversational momentum** - быстрые ответы во время активного диалога
+
+### Фаза 3: Advanced
+5. **Read tracker** - читать сообщения без мгновенного ответа
+6. **Sticker bank** - использовать стикеры вместо текста иногда
+7. **Quiet hours** - сниженная активность ночью
+8. **Proactive check-ins** - спросить как дела (осторожно)
+
+---
+
+## 🧪 Как протестировать
+
+1. Включите `typing_simulation: true` в config
+2. Отправьте короткое сообщение ("привет") → быстрый ответ
+3. Отправьте длинный вопрос → более долгое "думание"
+4. В группе ответы должны быть быстрее чем в DM
+5. Простые "ок" иногда приходят без typing indicator
+
+## 📊 Метрики
+
+До улучшений:
+- Фиксированная задержка `reply_delay_ms`
+- Все сообщения одинаковые
+- Не учитывается контекст (DM/group)
+
+После:
+- ✅ Динамические задержки (0.5s - 8s)
+- ✅ Зависит от длины ответа
+- ✅ Группы в 1.7× быстрее
+- ✅ Простые ответы могут быть мгновенными
+- ✅ ±30% variance для естественности
+
+---
+
+_Создано: 2025-01-XX_
+_Автор: Claude (Kiro)_
diff --git a/HUMANIZATION_ROADMAP.md b/HUMANIZATION_ROADMAP.md
diff --git a/README.md b/README.md
@@ -15,7 +15,7 @@
 
 ---
 
-<p align="center">Teleton is an autonomous AI agent platform that operates as a real Telegram user account or a Telegram Bot. It thinks through an agentic loop with tool calling, remembers conversations across sessions with hybrid RAG, and natively integrates the TON blockchain: send crypto, swap on DEXs, bid on domains, verify payments - all from a chat message. It can schedule tasks to run autonomously at any time. It ships with 131 always-registered tools plus 5 optional system tools, supports 16 LLM providers, and exposes a Plugin SDK so you can build your own tools on top of the platform.</p>
+<p align="center">Teleton is an autonomous AI agent platform that operates as a real Telegram user account or a Telegram Bot. It thinks through an agentic loop with tool calling, remembers conversations across sessions with hybrid RAG, and natively integrates the TON blockchain: verify payments and query on-chain state - all from a chat message. It can schedule tasks to run autonomously at any time. It ships with 102 always-registered tools plus 5 optional system tools, supports 16 LLM providers, and exposes a Plugin SDK so you can build your own tools on top of the platform.</p>
 
 ### Key Highlights
 
@@ -496,17 +496,13 @@ src/
 ├── agent/                  # Core agent runtime
 │   ├── runtime.ts          # Budgeted agentic loop, tool calling, masking, compaction
 │   ├── client.ts           # Multi-provider LLM client
-│   └── tools/              # 131 base tools plus 5 optional system tools
-│       ├── register-all.ts # Central tool registration (9 categories)
+│   └── tools/              # 102 base tools plus 5 optional system tools
+│       ├── register-all.ts # Central tool registration (5 categories)
 │       ├── registry.ts     # Tool registry, scope filtering, provider limits
 │       ├── module-loader.ts    # Built-in module loading (TON Proxy + exec)
 │       ├── plugin-loader.ts    # External plugin discovery, validation, hot-reload
 │       ├── mcp-loader.ts       # MCP client (stdio/SSE), tool discovery, lifecycle
-│       ├── telegram/       # Telegram operations (86 tools)
-│       ├── ton/            # TON blockchain + jettons + DEX router (15 tools)
-│       ├── stonfi/         # STON.fi DEX (5 tools)
-│       ├── dedust/         # DeDust DEX (5 tools)
-│       ├── dns/            # TON DNS (8 tools)
+│       ├── telegram/       # Telegram operations (83 tools)
 │       ├── exec/           # System execution — YOLO mode (4 tools)
 │       ├── journal/        # Business journal (3 tools)
 │       └── workspace/      # File operations (6 tools)

diff --git a/bin/teleton.js b/bin/teleton.js
@@ -1,2 +1,13 @@
 #!/usr/bin/env node
+import { execFileSync } from "node:child_process";
+
+if (process.platform === "win32" && process.stdout.isTTY) {
+  try {
+    // chcp changes the attached console, so UTF-8 log output is not decoded as CP866/1251.
+    execFileSync("chcp.com", ["65001"], { stdio: "ignore" });
+  } catch {
+    // Logging still works on terminals where the console code page cannot be changed.
+  }
+}
+
 await import("../dist/cli/index.js");
diff --git a/config.example.yaml b/config.example.yaml
@@ -3,7 +3,7 @@
 # Or run: teleton setup
 
 agent:
-  provider: "anthropic"              # anthropic | codex | grok-build | openai | google | xai | groq | openrouter
+  provider: "anthropic" # anthropic | codex | grok-build | openai | google | xai | groq | openrouter
   api_key: "YOUR_API_KEY"
   model: "claude-haiku-4-5-20251001" # Model ID (varies by provider)
   # utility_model: "claude-3-5-haiku-20241022"  # Optional: cheap model for summarization
@@ -13,46 +13,64 @@ agent:
   max_agentic_iterations: 5
   session_reset_policy:
     daily_reset_enabled: true
-    daily_reset_hour: 4              # 0-23, hour of day to reset sessions
+    daily_reset_hour: 4 # 0-23, hour of day to reset sessions
     idle_expiry_enabled: true
-    idle_expiry_minutes: 1440        # 24 hours
+    idle_expiry_minutes: 1440 # 24 hours
 
 telegram:
-  mode: "user"                       # "user" (default) or "bot" — user requires phone auth, bot requires bot_token
-  api_id: 0                          # From https://my.telegram.org/apps
+  mode: "user" # "user" (default) or "bot" — user requires phone auth, bot requires bot_token
+  api_id: 0 # From https://my.telegram.org/apps
   api_hash: "YOUR_API_HASH"
-  phone: "+1234567890"               # Phone number linked to Telegram account
-  session_name: "teleton_session"    # GramJS session file name
-  session_path: "~/.teleton"         # Directory for session files
-  dm_policy: "open"                  # open | allowlist | admin-only | disabled
-  allow_from: []                     # Telegram user IDs (for allowlist policy)
-  group_policy: "open"               # open | allowlist | disabled
-  group_allow_from: []               # Group chat IDs (for allowlist group policy)
-  require_mention: true              # In groups, only respond when mentioned
+  phone: "+1234567890" # Phone number linked to Telegram account
+  session_name: "teleton_session" # GramJS session file name
+  session_path: "~/.teleton" # Directory for session files
+  dm_policy: "open" # open | allowlist | admin-only | disabled
+  allow_from: [] # Telegram user IDs (for allowlist policy)
+  group_policy: "open" # open | allowlist | disabled
+  group_allow_from: [] # Group chat IDs (for allowlist group policy)
+  require_mention: true # In groups, only respond when mentioned
   # max_message_length: 4096         # MTProto limit, rarely needs changing
-  typing_simulation: true            # Show "typing..." while processing
+  typing_simulation: true # Show "typing..." while processing
+  reaction_events: true # Let the bot receive reactions to its messages
   rate_limit_messages_per_second: 1.0
   rate_limit_groups_per_minute: 20
-  admin_ids: [123456789]             # Your Telegram user ID (from @userinfobot)
+  admin_ids: [123456789] # Your Telegram user ID (from @userinfobot)
   # agent_channel: null              # Channel username or ID for announcements
   owner_name: "Your Name"
   owner_username: "your_username"
   # owner_id: 123456789              # Owner's Telegram user ID
-  debounce_ms: 1500                  # Group message batching delay (0 = disabled)
+  debounce_ms: 1500 # Group message batching delay (0 = disabled)
+  dm_debounce_ms: 1000 # Direct-message batching window: rapid DMs are merged into one context-aware reply (0 = disabled)
 
   # Optional in user mode: plugin inline cards and callback handling
   # bot_token: "123456:ABC-DEF..."   # From @BotFather
   # bot_username: "your_agent_bot"
 
+# Optional: periodic heartbeat and proactive checks
+heartbeat:
+  enabled: true
+  interval_ms: 3600000 # Every 60 min
+  # startup_prompt: "Greet me once at startup." # One-time prompt on first tick
+  min_interval_between_replies_ms: 30000 # Min time between two heartbeat replies to the same chat (0 = off)
+  reply_delay_ms: 1000 # Human-like delay before sending a heartbeat reply (0 = none)
+
+# Optional: Enable TON blockchain features (wallet, DEX, DNS, jettons, NFTs)
+# ton_features:
+#   enabled: false                    # Set to true to enable TON SDK in plugins
+
 # Optional: TonAPI key for higher rate limits
 # tonapi_key: "YOUR_TONAPI_KEY"      # From @tonapi_bot
 
 # Optional: Tavily API key for web search/fetch tools
 # tavily_api_key: "YOUR_TAVILY_KEY"  # From https://tavily.com
 
+# Optional: MiniMax TTS keys for telegram_send_voice with ttsProvider: minimax
+# MINIMAX_API_KEY: sk-cp-WJx00mS7JhhsX4PxDqVBBdHZAihV3_KEbbtnM-MRC1uNBYFfbhT3X9fPEKSd1etdPyk6BXPL3N3-hN83hKEitfSLiOodkp8k_IQQruxcqnvNAL_OUuyoGTA   # From https://platform.minimax.io/
+# MINIMAX_GROUP_ID: "YOUR_MINIMAX_GROUP_ID" # Account GroupId (see platform console)
+
 # Embedding provider for RAG vector search
 embedding:
-  provider: "local"                  # local | anthropic | none
+  provider: "local" # local | anthropic | none
   # model: "Xenova/all-MiniLM-L6-v2" # Model override (auto-detected per provider)
   #
   # local    = ONNX runtime, runs entirely on CPU, no API calls needed
@@ -61,14 +79,15 @@ embedding:
 
 # Optional: WebUI dashboard (disabled by default for security)
 webui:
-  enabled: false                      # Enable web dashboard
-  port: 7777                          # HTTP server port
-  host: "127.0.0.1"                   # Localhost only (secure)
+  enabled: false # Enable web dashboard
+  port: 7777 # HTTP server port
+  host: "127.0.0.1" # Localhost only (secure)
   # auth_token: "..."                 # Auto-generated if omitted
   cors_origins:
-    - "http://localhost:5173"         # Vite dev server
-    - "http://localhost:7777"         # Production server
-  log_requests: false                 # Log HTTP requests
+    - "http://localhost:5173" # Vite dev server
+    - "http://localhost:7777" # Production server
+  log_requests: false # Log HTTP requests
+
 
 # Tool RAG: semantic tool selection per message (ENABLED by default)
 # Reduces tool definitions sent to the LLM from 120+ to ~25 most relevant.
@@ -106,8 +125,35 @@ webui:
 #     remote:
 #       url: http://localhost:3001/mcp
 
+# Optional: Human-like behavior enhancements
+# Makes the agent behave more like a real person
+humanization:
+  enabled: true
+  reply_probability:
+    dm_base: 0.85            # 85% chance to reply in DMs
+    group_mentioned: 0.50    # 50% chance when mentioned in group
+    group_replied_to_us: 0.90 # 90% if someone replied to our message
+    group_unmentioned: 0.05  # 5% without mention (smart silence)
+    min_interval_ms: 3000    # Don't reply faster than 3s in same chat
+    high_activity_threshold: 5  # Messages/min threshold for "busy chat"
+    high_activity_multiplier: 0.3  # Reduce replies 70% in busy groups
+  time_of_day:
+    enabled: true
+    quiet_hours_start: 23    # 11 PM
+    quiet_hours_end: 6       # 6 AM
+    timezone_offset_minutes: 180  # UTC+3 (Moscow). Adjust for your zone
+  writing_style:
+    typo_probability: 0.03       # 3% chance of simulating a typo
+    edit_after_send_probability: 0.05  # 5% chance of edit after send
+    max_edit_delay_ms: 15000     # Max delay for edit simulation
+    staging_probability: 0.03    # 3% show typing but don't send
+    emoji_enabled: true
+    fillers_enabled: true
+
 # Per-plugin configuration (key = plugin name)
 # plugins:
 #   casino:
 #     min_bet: 0.1
 #     cooldown_seconds: 30
+# Spotify search and track metadata (optional; create an app at developer.spotify.com)
+# Set SPOTIFY_CLIENT_ID and SPOTIFY_CLIENT_SECRET in the environment.